Atlanta/Magellan ...Redux

[Steve B]

Thanks Nick

not quite sure where our discussion stands at this point?
i am still interested in the Atlanta-Magellan thingy
not sure why the Senator has crept into the Discussion?

Anyway i guess we can go over all of this when your tests are all completed

Bit Confused Regards
Steve

[spacious_mind]

Here are the results after 5 Test Games. Selective Ponder On 30 Seconds per move:

=========================TEST GAME 1===============================


------ATLANTA------ -----MAGELLAN------ ------SENATOR------
7. d5 2.5 7. d5 2.5 7. d5 2.5
8. Bg5 1.6X 8. Be2 0.0 8. Be2 0.0
9. Bxf6 1.0 9. Bxf6 1.0 9. Bxf6 1.0
10. Qd2 2.6 10. Qd2 2.6 10. Qd2 2.6
11. Rd1 3.0 11. Rd1 3.0 11. Rd1 3.0
12. Rd1 2.4 12. Rd1 2.4 12. Rd1 2.4
13, h3 2.7 13, h3 2.7 13, h3 2.7
14. Qc2 2.0 14. Qc2 2.0 14. Qc2 2.0
15. h3 1.0 15. h3 1.0 15. h3 1.0
16. h3 1.4 16. h3 1.4 16. h3 1.4
17. h3 3.0 17. h3 3.0 17. h3 3.0
18. h3 3.0 18. h3 3.0 18. h3 3.0
19. Na4 3.0 19. Na4 3.0 19. Na4 3.0
20. Qxa4 3.0 20. Qxa4 3.0 20. Qxa4 3.0
21. Rc4 4.0X 21. Bg4 2.3 21. Bg4 2.3
22. b4 4.0 22. b4 4.0 22. b4 4.0
23. Rxc5 4.4 23. Rxc5 4.4 23. Rxc5 4.4
24. dxe6 4.0 24. dxe6 4.0 24. dxe6 4.0
25. Rxd6 3.2 25. Rxd6 3.2 25. Rxd6 3.2
26. Rc8+ 4.0 26. Rc8+ 4.0 26. Rc8+ 4.0
27. Qc7+ 2.0 27. Qc7+ 2.0 27. Qc7+ 2.0
28. Rxh8 4.0X 28. Bg4+ 4.0 28. Bg4+ 4.0
29. Bxf5+ 3.0 29. Bxf5+ 3.0 29. Bxf5+ 3.0
30. Qc5+ 3.0 30. Qc5+ 3.0 30. Qc5+ 3.0

Magellan comparison to Atlanta = 3 deviations from 24 = 87.50% Match or 12.50% Mismatch
Senator comparison to Atlanta = 3 deviations from 24 = 87.50% Match or 12.50% Mismatch
Senator comparison to Magellan = 0 deviation from 24 = 100.00% Match or 0% Mismatch


=========================TEST GAME 2===============================

------ATLANTA------ -----MAGELLAN------ ------SENATOR------
6. … e5 2.5 6. … e5 2.5 6. … e5 2.5
7. … Be7 1.3 7. … Be7 1.3 7. … Be7 1.3
8. … Nxd5 3.0 8. … Nxd5 3.0 8. … Nxd5 3.0
9. … Be6 4.0 9. … Be6 4.0 9. … Be6 4.0
10. … Qd7 4.0X 10. … Qh4 1.6 10. … Qh4 1.6
11. … Nxe3 4.0 11. … Nxe3 4.0 11. … Nxe3 4.0
12. … Qd7 1.0 12. … Qd7 1.0 12. … Qd7 1.0
13. … f5 4.0 13. … f5 4.0 13. … f5 4.0
14. … Qc8 2.0 14. … Qc8 2.0 14. … Qc8 2.0
15. … Qxe6 1.0 15. … Qxe6 1.0 15. … Qxe6 1.0
16. … f4 0.0 16. … f4 0.0 16. … f4 0.0
17. … f4 0.0 17. … f4 0.0 17. … f4 0.0
18. … Kh8 2.0 18. … Kh8 2.0 18. … Kh8 2.0
19. … Qe7 1.5X 19. … Rfd8 0.0 19. … Rfd8 0.0
20. … b5 1.2X 20. … Qf6 4.0 20. … Qf6 4.0
21. … Rfd8 0.0 21. … Rfd8 0.0 21. … Rfd8 0.0
22. … Bxc3 0.0 22. … Bxc3 0.0 22. … Bxc3 0.0
23. … fxg3 0.0 23. … fxg3 0.0 23. … fxg3 0.0
24. … exf2+ 4.0 24. … exf2+ 4.0 24. … exf2+ 4.0
25. … Rxf8 3.0 25. … Rxf8 3.0 25. … Rxf8 3.0
26. … fxg3 3.0 26. … fxg3 3.0 26. … fxg3 3.0
27. … gxh2 2.0 27. … gxh2 2.0 27. … gxh2 2.0
28. … Qg6+ 1.0 28. … Qg6+ 1.0 28. … Qg6+ 1.0

Magellan comparison to Atlanta = 3 deviations from 23 = 86.96% Match or 13.04% Mismatch
Senator comparison to Atlanta = 3 deviations from 23 = 86.96% Match or 13.04% Mismatch
Senator comparison to Magellan = 0 deviation from 23 = 100.00% Match or 0% Mismatch

=========================TEST GAME 3===============================

------ATLANTA------ -----MAGELLAN------ ------SENATOR------
10. … e6 0.0X 10. … Bd7 3.5 10. … Bd7 3.5
11. … Bd7 3.8 11. … Bd7 3.8 11. … Bd7 3.8
12. … Ne5 3.5 12. … Ne5 3.5 12. … Ne5 3.5
13. … Qc8 3.5 13. … Qc8 3.5 13. … Qc8 3.5
14. … Nac4 3.5 14. … Nac4 3.5 14. … Nac4 3.5
15. … Ng4+ 1.2 15. … Ng4+ 1.2 15. … Ng4+ 1.2
16. … Ng4+ 0.0 16. … Ng4+ 0.0 16. … Ng4+ 0.0
17. … Qc4 4.0 17. … Qc4 4.0 17. … Qc4 4.0
18. … Rac8 3.8X 18. … Rfc8 4.0 18. … Rfc8 4.0
19. … Rc7 2.9X 19. … Qb5 0.0 19. … Qb5 0.0
20. … Qxd2 4.0 20. … Qxd2 4.0 20. … Qxd2 4.0
21. … b5 1.0 21. … b5 1.0 21. … b5 1.0
22. … Bc2 3.8 22. … Bc2 3.8 22. … Bc2 3.8
23. … Nxf3+ 3.4 23. … Nxf3+ 3.4 23. … Nxf3+ 3.4
24. ...Rac8 4.0 24. ...Rac8 4.0 24. ...Rac8 4.0
25. … Qxb4 3.0X 25. … b5 0.0 25. … b5 0.0
26. … a5 4.0 26. … a5 4.0 26. … a5 4.0
27. … Qb6 3.5 27. … Qb6 3.5 27. … Qb6 3.5
28. … Rc3 3.5 28. … Rc3 3.5 28. … Rc3 3.5
29. … R8c4 1.5 29. … R8c4 1.5 29. … R8c4 1.5
30. … Qf2+ 3.5 30. … Qf2+ 3.5 30. … Qf2+ 3.5

Magellan comparison to Atlanta = 4 deviations from 21 = 80.95% Match or 19.05% Mismatch
Senator comparison to Atlanta = 4 deviations from 21 = 80.95% Match or 19.05% Mismatch
Senator comparison to Magellan = 0 deviation from 23 = 100.00% Match or 0% Mismatch

=========================TEST GAME 4===============================

------ATLANTA------ -----MAGELLAN------ ------SENATOR------
16. Ne3 4.0 16. Ne3 4.0 16. Ne3 4.0
17. Nd5 4.0 17. Nd5 4.0 17. Nd5 4.0
18. Nd5 3.0 18. Nd5 3.0 18. Nd5 3.0
19. Bd2 1.0 19. Bd2 1.0 19. Bd2 1.0
20. d5 1.8 20. d5 1.8 20. d5 1.8
21. Bg5 4.0 21. Bg5 4.0 21. Bg5 4.0
22. d5 0.0 22. d5 0.0 22. d5 0.0
23. d5 4.0X 23. dxe5 1.6 23. d5 4.0X
24. d5 3.7 24. d5 3.7 24. d5 3.7
25. Nd5 4.0 25. Nd5 4.0 25. Nd5 4.0
26. a5 4.0 26. a5 4.0 26. a5 4.0
27. a5 4.0 27. a5 4.0 27. a5 4.0
28. Bxb4 1.9X 28. Nxb4 1.6 28. Nxb4 1.6
29. Ba3 2.1 29. Ba3 2.1 29. Ba3 2.1
30. Qe2 0.0 30. Qe2 0.0 30. Qe2 0.0
31. Nxe5 4.0 31. Nxe5 4.0 31. Nxe5 4.0
32. Rb4 4.0 32. Rb4 4.0 32. Rb4 4.0
33. Rxb8 4.0 33. Rxb8 4.0 33. Rxb8 4.0
34. Qxa6 4.0 34. Qxa6 4.0 34. Qxa6 4.0
35. Rc8 4.0 35. Rc8 4.0 35. Rc8 4.0

Magellan comparison to Atlanta = 2 deviations from 20 = 90.00% Match or 10.00% Mismatch
Senator comparison to Atlanta = 1 deviation from 20 = 95.00% Match or 5.00% Mismatch
Senator comparison to Magellan = 1 deviation from 20 = 95.00% Match or 5.00% Mismatch

TOTAL AFTER 5 TEST GAMES

Magellan comparison to Atlanta = 19 deviations from 116 = 83.62% Match or 17.48% Mismatch
Senator comparison to Atlanta = 17 deviation from 116 = 85.34% Match or 14.66% Mismatch
Senator comparison to Magellan = 2 deviation from 116 = 98.28% Match or 1.72% Mismatch

Best regards,

Nick

[Steve B]

Magellan comparison to Atlanta = 19 deviations from 116 = 83.62% Match or 17.48% Mismatch

I guess the test is concluded?

Well we have a 83.62% Match..with pondering On
very close to my threshold of 85-90% to be considered a clone but
still..just under it

i would like to see at least some sort of Clarification for the Magellan listing on the Wiki mentioning the existence of 16 Mhz Magellans and these so-called Magellan(Senators) which are influencing the Wiki rating
probably not going to happen though

anyway...
Thank You very much for these tests and for the detour in your other tests regarding strength Vs Humans

Best Regards
Steve

[spacious_mind]

Hi Steve,

It is not a detour since I want to rate all my computers anyway. I will continue with them throughout the 16 Tests. Test game 4 had a 90% match so it is likely that in some games a 100% match will be found.

The validation comes through others trying the same test and see what their computers show.

At the moment I cannot say that a 16 MHz exists or that Atlanta and Magellan are exact clones because all the test I did 5 years ago and all the tests that I have done with my machines today indicate that there is a difference.

All my 3 computers are 20 MHz.

Best regards,

Nick

[Theo]

Hi Nick,

where can we read about the exact algorithm for your test?

So far I understood the following:

1. Computer is set to 30sec/move
2. PONDER = ON ?
3. Forward the game to the starting position

4. Let the computer compute the next move
5. If move = game move, wait 30 Secs (?) and make opponent move, then go to step 4.
6. Else take the move back, make the correct move and the opponent move, then go to step 4.

7. Note the points according to the moves found.

Is that correct?

I am reducing my collection, had to put them all away from my mother's place where I collected them since the 90ies . And my dilemma is, should I keep a Milano Pro additionally to my Magellan or are they too similar. So far I have the feeling though they both are related, the Milano Pro plays a much more exciting chess. I would like to run it through your test, but I noticed I am not sure about some points.

Kind Regards,
Theo

[spacious_mind]

Hi Nick,

where can we read about the exact algorithm for your test?

So far I understood the following:

1. Computer is set to 30sec/move
2. PONDER = ON ?
3. Forward the game to the starting position

4. Let the computer compute the next move
5. If move = game move, wait 30 Secs (?) and make opponent move, then go to step 4.
6. Else take the move back, make the correct move and the opponent move, then go to step 4.

7. Note the points according to the moves found.

Is that correct?

I am reducing my collection, had to put them all away from my mother's place where I collected them since the 90ies . And my dilemma is, should I keep a Milano Pro additionally to my Magellan or are they too similar. So far I have the feeling though they both are related, the Milano Pro plays a much more exciting chess. I would like to run it through your test, but I noticed I am not sure about some points.

Kind Regards,
Theo

Hi Theo,

http://spacious-mind.com/forum_reports/ ... _final.ods

Above is the Spreadsheet that is being used. So far only 30 seconds per move has been tried.

Magellan should be stronger than MP so I personally would keep Magellan. An example of how I do the evaluation criteria is on the first page here:

http://hiarcs.net/forums/viewtopic.php? ... sc&start=0

To test them just use PS0 with MP and set them both up with Selective. Then just enter the move they played on the spreadsheet and the points next to the move. If the move is not listed then there are no points for the move. The spreadsheet will do the calculations for you.

Regards

[Theo]

Hi Nick,

thanks for the rapid response! I understand the evaluation criteria and also I know where to find the games.

What I am still uncertain about is the precise steps of execution of this test.

So pondering is ON per default? And if its ON, do I wait 30secs after the (right) move to give the machine a chance for a ponder hit?

Why is the level 30sec/move, shouldn't the time needed for the test moves of one whole game been evaluated as well? Because computers have different time management.

I'd prefer to use the fixed amount of 30sec per move so I would set the machine to "Analyze"-level and then manually stop it after 30 secs.

Kind Regards,
Theo

[spacious_mind]

Hi Nick,

thanks for the rapid response! I understand the evaluation criteria and also I know where to find the games.

What I am still uncertain about is the precise steps of execution of this test.

So pondering is ON per default? And if its ON, do I wait 30secs after the (right) move to give the machine a chance for a ponder hit?

Why is the level 30sec/move, shouldn't the time needed for the test moves of one whole game been evaluated as well? Because computers have different time management.

I'd prefer to use the fixed amount of 30sec per move so I would set the machine to "Analyze"-level and then manually stop it after 30 secs.

Kind Regards,
Theo

Hi Theo,

I have just tried to simulate normal play conditions so 30 seconds per move ponder on or if you are playing 3 minutes per move then 40/2hrs or 3 minutes per move. There is a tab for 3 minutes per move but I have not started any games yet.

I normally let about 30 senconds pass before I enter the next move, but only if the computer guessed the correct one, If he didn't then it doesnt matter because the take back move makes the computer calculate anyway.

I see this test a comparison to normal play there I am not that interested in analysis mode.

Regards
Nick

[Reinfeld]

I'm an idiot. I just realized the significance of this argument.

If Nick's tests are right, then the Wiki ratings are thoroughly messed up. It appears the Magellan is a Senator and *not* an Atlanta.

A little earlier, Steve called the rating difference "very small." Larry's phrase was, "the persistent rating difference."

True, especially when you're talking about Magellan v. Atlanta. You can slice the ratings several ways, but they're pretty close across most categories. Aktiv rating:

Atlanta - 2267
Magellan - 2234

Averaging all ratings gets them even closer - 4 points apart.

So OK, not so much. But add the aktiv rating for Senator:

Atlanta - 2267
Magellan - 2234
Senator - 2139

Now - consider the NICK-ELO after 6 test games:

Atlanta - 2346
Senator - 2264
Magellan - 2252

That, friends, looks to me like a significant rating difference. Whether you buy Nick's numbers or Wiki-Elo, you're talking around 100 points.

Possibilities:

- Nick owns a souped-up Senator
- Nick owns a defective Magellan
- Nick owns a gelded 16 Mhz Magellan
- The Great Powers lied about the 20 Mhz Magellan
- The Great Powers lied about the 20 Mhz Senator

Fact: Atlanta is stronger than both by every measure.


- R.

[Theo]

Please note that on the Great .Info Forum some god of German engineering soldered a 512kB RAM into a Senator and, voilà, suddenly it was a Magellan with same solution times in the tests he performed.

[spacious_mind]

Please note that on the Great .Info Forum some god of German engineering soldered a 512kB RAM into a Senator and, voilà, suddenly it was a Magellan with same solution times in the tests he performed.

Yes, but mine is just a regular Senator. I had opened both up before I started the Test and the RAM slot is empty on a Senator, otherwise Magellan and Senator are both 100% identical inside the module from a design and layout perspective.

It could be that in these tests which are at 30 seconds per move the 512K RAM does not get a chance to kick in. Look at Micha's example with Risc 2500 128K vs 2 MB he also indicated that in active chess the ELO difference is almost non existent.

If you look at game 6 of my Tests, Atlanta, Magellan and Senator played the same moves 100%. So they are all identical. I just think that with Atlanta there are some key evaluations that were fine tuned and are missing in Senator and Magellan and this is why I think that Atlanta in some situation plays a different move and most often when it happens plays a better move than Magellan or Senator. In my opinion Info's ELO difference of about 40 is about right!

Regards,

[Theo]

Nick,

.info ELO says Atlanta is better at rapid chess, while Magellan is much better (46 points!!) at longer time control. Don't forget that

Kind Regards,
Theo

[Steve B]

Please note that on the Great .Info Forum some god of German engineering soldered a 512kB RAM into a Senator and, voilà, suddenly it was a Magellan with same solution times in the tests he performed.

not sure what the confusion is here
is there any doubt that the Senator is the same exact computer as the Magellan except for the Ram?
it was sold and advertised this way and all of the literature confirms this

Not Even An issue Regards
Steve

[Steve B]

Nick,

.info ELO says Atlanta is better at rapid chess, while Magellan is much better (46 points!!) at longer time control. Don't forget that

Kind Regards,
Theo

your forgot to use the word Great when discussing .INFO
I can edit your post to correct this glaring error if you wish

Always Helpful Regards
Steve

[Steve B]

If you look at game 6 of my Tests, Atlanta, Magellan and Senator played the same moves 100%. So they are all identical.

and they all have the same bug as evidenced by the test position you asked me to try out
so I am a bit confused
you are saying they are identical and yet you still feel they deserve a separate listing on Wiki
the 40 elo wiki difference can be due to several factors
the ones your mentioned,pondering during game play,random factor ..etc..etc

my guess is if you took two of the same computer..any computer..
lets say the Saitek Sparc..two identical Sparcs
and you just labeled them computer 1...computer 2
and then ran them against different opponents(as the Atlanta and Magellan were on Wiki) you would get a small difference in rating
this would be true at any Time Control

we both said to wait until your done with all of your games to draw any final conclusions but...
if when done if the match ratio between Magellan and Atlanta is at or near 90% ...
will you still hold the position that they deserve a separate listing on Wiki ?

if your answer is still YES .then I guess we don't need to debate this any longer as it would make no sense to from my perspective

Best Regards
Steve